Upload config.json
#3
by
thanhnhan29
- opened
- config.json +7 -24
config.json
CHANGED
|
@@ -1,8 +1,6 @@
|
|
| 1 |
{
|
| 2 |
"_name_or_path": "/tmp/iopath_cache/manifold_cache/tree/users/shenx/finetune/09281004-cambrian_llama3_2_t576_ov",
|
| 3 |
-
"architectures": [
|
| 4 |
-
"CambrianLlamaForCausalLM"
|
| 5 |
-
],
|
| 6 |
"attention_bias": false,
|
| 7 |
"attention_dropout": 0.0,
|
| 8 |
"bos_token_id": 128000,
|
|
@@ -11,11 +9,7 @@
|
|
| 11 |
"connector_only": true,
|
| 12 |
"dino_threshold": 0.83,
|
| 13 |
"drop_threshold": 0.8,
|
| 14 |
-
"eos_token_id": [
|
| 15 |
-
128001,
|
| 16 |
-
128008,
|
| 17 |
-
128009
|
| 18 |
-
],
|
| 19 |
"frame_pos": false,
|
| 20 |
"freeze_mm_mlp_adapter": false,
|
| 21 |
"hidden_act": "silu",
|
|
@@ -44,10 +38,7 @@
|
|
| 44 |
"siglip/CLIP-ViT-SO400M-14-384",
|
| 45 |
"facebook/dinov2-giant-res378"
|
| 46 |
],
|
| 47 |
-
"mm_vision_tower_aux_token_len_list": [
|
| 48 |
-
576,
|
| 49 |
-
576
|
| 50 |
-
],
|
| 51 |
"mm_vision_tower_lr": null,
|
| 52 |
"model_type": "cambrian_llama",
|
| 53 |
"num_attention_heads": 24,
|
|
@@ -56,16 +47,11 @@
|
|
| 56 |
"num_of_vision_sampler_layers": 10,
|
| 57 |
"num_query_group": 1,
|
| 58 |
"pretraining_tp": 1,
|
| 59 |
-
"query_num_list": [
|
| 60 |
-
|
| 61 |
-
],
|
| 62 |
-
"rms_norm_eps": 1e-05,
|
| 63 |
"rope_scaling": {
|
| 64 |
"factor": 32.0,
|
| 65 |
-
"
|
| 66 |
-
"low_freq_factor": 1.0,
|
| 67 |
-
"original_max_position_embeddings": 8192,
|
| 68 |
-
"rope_type": "llama3"
|
| 69 |
},
|
| 70 |
"rope_theta": 500000.0,
|
| 71 |
"spmd_debug": null,
|
|
@@ -83,9 +69,6 @@
|
|
| 83 |
"use_cache": false,
|
| 84 |
"use_mm_proj": true,
|
| 85 |
"vision_hidden_size": 1024,
|
| 86 |
-
"vision_tower_aux_token_len_list": [
|
| 87 |
-
576,
|
| 88 |
-
576
|
| 89 |
-
],
|
| 90 |
"vocab_size": 128256
|
| 91 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"_name_or_path": "/tmp/iopath_cache/manifold_cache/tree/users/shenx/finetune/09281004-cambrian_llama3_2_t576_ov",
|
| 3 |
+
"architectures": ["CambrianLlamaForCausalLM"],
|
|
|
|
|
|
|
| 4 |
"attention_bias": false,
|
| 5 |
"attention_dropout": 0.0,
|
| 6 |
"bos_token_id": 128000,
|
|
|
|
| 9 |
"connector_only": true,
|
| 10 |
"dino_threshold": 0.83,
|
| 11 |
"drop_threshold": 0.8,
|
| 12 |
+
"eos_token_id": [128001, 128008, 128009],
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
"frame_pos": false,
|
| 14 |
"freeze_mm_mlp_adapter": false,
|
| 15 |
"hidden_act": "silu",
|
|
|
|
| 38 |
"siglip/CLIP-ViT-SO400M-14-384",
|
| 39 |
"facebook/dinov2-giant-res378"
|
| 40 |
],
|
| 41 |
+
"mm_vision_tower_aux_token_len_list": [576, 576],
|
|
|
|
|
|
|
|
|
|
| 42 |
"mm_vision_tower_lr": null,
|
| 43 |
"model_type": "cambrian_llama",
|
| 44 |
"num_attention_heads": 24,
|
|
|
|
| 47 |
"num_of_vision_sampler_layers": 10,
|
| 48 |
"num_query_group": 1,
|
| 49 |
"pretraining_tp": 1,
|
| 50 |
+
"query_num_list": [144],
|
| 51 |
+
"rms_norm_eps": 1e-5,
|
|
|
|
|
|
|
| 52 |
"rope_scaling": {
|
| 53 |
"factor": 32.0,
|
| 54 |
+
"type": "llama3"
|
|
|
|
|
|
|
|
|
|
| 55 |
},
|
| 56 |
"rope_theta": 500000.0,
|
| 57 |
"spmd_debug": null,
|
|
|
|
| 69 |
"use_cache": false,
|
| 70 |
"use_mm_proj": true,
|
| 71 |
"vision_hidden_size": 1024,
|
| 72 |
+
"vision_tower_aux_token_len_list": [576, 576],
|
|
|
|
|
|
|
|
|
|
| 73 |
"vocab_size": 128256
|
| 74 |
}
|